“…These histograms show that majority of time series subsequences considered are less than 1,000 datapoints, and more importantly, the best value for w is rarely above 10%. Almost all uses of FastDTW fall into Case A [7], [8], [9], [10], [11], [12], [13], [14], [15], [16], [17], [18], and in every case the researchers using FastDTW would have been better off using classic cDTW, which would have been much faster, and exact.…”